UAV-enabled Edge Computing for Virtual Reality

5G communication promotes the development of VR (Virtual Reality) applications, providing users with immersive experiences. To accomplish VR tasks with large computation and low delay demands, an unmanned aerial vehicle (UAV)-enabled MEC (Mobile Edge Computing) method is proposed to assist VR devices in the rendering process. Under the constraints imposed by the VR characteristics and the device energy, the UAV flight trajectory and the VR rendering mode are jointly optimized to maximize the rendering completion rate of the VR tasks. This problem is modeled as a Markov decision process. To find the optimal policy, a UAV aided rendering algorithm is proposed in the framework of deep reinforcement learning. Specifically, the TD3 (Twin Delayed Deep Deterministic Policy Gradient) algorithm is applied to schedule the UAV trajectory and VR rendering mode to meet the requirements of the randomly arriving VR tasks as much as possible. Simulation results show that the proposed method outperforms baseline strategies in both the rendering completion rate and the convergence speed.